A theory-driven learning analytic model for detecting students at risk in higher education

Abstract

Business intelligence (BI) and analytic solutions originated from learning analytics (LA), which has emerged in the education sector due to the success of data mining models in businesses. However, the application of BI models in analytics for at-risk students (those failing academically or facing challenges that may hinder their completion of studies) is currently unclear in the global higher education context. LA is being tested and implemented in some higher education institutions (HEI) worldwide to enhance learning and teaching through monitoring students’ interactions and success in fully online and blended courses. Furthermore, most studies on LA are data-driven and lack theoretical foundations. This study, from which the findings in this article are derived, is based on Tinto’s longitudinal model of dropout. This model is used to select dropout conditions and extract data from institutional information systems and student learning data, with the goal of improving the identification of at-risk students and providing real-time interventions. Through an inductive analysis of literature, this article explores how theoretical frameworks can be applied in analytics for at-risk students, with a focus on predictive modeling. As a result, a modified theoretical model based on Tinto’s longitudinal model of dropout is presented. This modified model aims to demonstrate the potential of information systems and student learning data in indigenous HEI, providing a learning analytic approach that universities can use to identify students at risk of dropping out.